Gary Marcus on Ban superintelligence development until…

Gary Marcus

Professor of Psychology and Neural Science

My opinion is that the moratorium that we should focus on is actually deployment until we have good safety cases. I don't know that we need to pause that particular project, but I do think it's emphasis on focusing more on AI safety, on trustworthy, reliable AI is exactly right. [...] I would agree. And, and I don't think it's a realistic thing in the world. The reason I personally signed the letter was to call attention to how serious the problems were and to emphasize spending more of our efforts on trustworthy and safe ai rather than just making a bigger version of something we already know to be unreliable.

AI Verified source (May 16, 2023)

11mo ago

Policy proposals and claims

Ban superintelligence development until safety consensus is reached

votes Against

Statement relation verification history AI Verified Report this

AI Verified The quote clearly implies opposition to the full statement: the author says the moratorium should be on deployment, not development, says "I don't know that we need to pause that particular project," and calls such a pause unrealistic. That establishes a stance against banning superintelligence development until safety consensus. · YouCongress gpt-5.4-2026-03-05 · 3mo ago

Vote inference verification history AI Verified Report this

AI Verified The quote points away from a development ban: the author says the moratorium should be on "deployment until we have good safety cases" and "I don't know that we need to pause that particular project," also calling such a pause "not a realistic thing in the world." · YouCongress gpt-5.4-2026-03-05 · 3mo ago

Quote authenticity verification history

Report this

AI Verified The Tech Policy Press transcript page dated May 16, 2023 attributes both quoted passages to Gary Marcus, and the source URL contains the matching wording at the relevant points in the hearing transcript. The stored quote is an accurate excerpt with an allowed internal omission ([...]); it also simply stops before a trailing "Yeah." in the second passage, which does not materially alter the text. The stored author, date, and source URL are consistent with the source. ([techpolicy.press](https://www.techpolicy.press/transcript-senate-judiciary-subcommittee-hearing-on-oversight-of-ai/)) · YouCongress gpt-5.4-2026-03-05 · 3mo ago

Disputed The remarks are authentic and correctly attributed to Gary Marcus in the May 16, 2023 Senate hearing, but the submitted quote is not strictly verbatim to one reliable source. The official Congress transcript reads “its emphasis” and later “And I don’t think…,” while the Tech Policy Press rush transcript has “it’s emphasis” and “And, and I don’t think…”. So the quote as given appears to blend transcript variants rather than exactly matching a single source. ([congress.gov](https://www.congress.gov/event/118th-congress/senate-event/LC71543/text)) · YouCongress gpt-5.4-2026-03-05 · 3mo ago

AI Verified Source URL (techpolicy.press transcript of May 16, 2023 Senate Judiciary Subcommittee 'Oversight of AI' hearing) returns 403 to WebFetch but Marcus's testimony at this hearing is extensively documented. Search results confirm Marcus's substantive position matches the quote: (a) he emphasized deployment-stage safety cases and a nimble monitoring agency for post-deployment review; (b) he proposed FDA/aerospace-style certification rather than a blanket development pause; (c) he was a signatory of the FLI pause letter and explained his motivation was to highlight the seriousness of risks and shift focus toward trustworthy AI rather than literally halt development. Author attribution to Gary Marcus (NYU emeritus, Professor of Psychology and Neural Science) is correct. Year 2023 matches the hearing date. Vote 'against' on statement #379 ('Ban superintelligence development until safety consensus is reached') aligns with the quote's content: Marcus explicitly says 'I don't know that we need to pause that particular project' and 'I don't think it's a realistic thing in the world,' favoring deployment-moratorium and safety cases instead. Verified. · Hector Perez Arenas claude-opus-4-7 · 4mo ago

replying to Gary Marcus

Other opinions from Gary Marcus

See all

Gary Marcus votes Against

AGI could quickly lead to superintelligence

Plot the same improvements on a linear y-axis and you get steady incremental gains, impressive but not the hockey stick that triggered the panic. [...] We don't have evidence that Mythos is actually an important step towards broad superintelligence.

Disputed source (2026)

added 4mo ago

Gary Marcus votes For

Ban autonomous lethal weapons

The White House and people at the top of the Pentagon seem to treat it as magic, whereas those who are actually building it realize that it hallucinates, it makes mistakes, stupid errors of reasoning. And if you put that into production, into practic...

more Unverifiable source (2026)

added 5mo ago

Gary Marcus votes Against